Development and external validation of a clinical prediction model for functional impairment after intracranial tumor surgery

Machine Learning in Neurosurgery: Toward Complex Inputs, Actionable Predictions, and Generalizable Translations

Machine learning can predict neurosurgical diagnosis and outcomes, power imaging analysis, and perform robotic navigation and tumor labeling. State-of-the-art models can reconstruct and generate images, predict surgical events from video, and assist in intraoperative decision-making. In this review, we will detail the neurosurgical applications of machine learning, ranging from simple to advanced models, and their potential to transform patient care. As machine learning techniques, outputs, and methods become increasingly complex, their performance is often more impactful yet increasingly difficult to evaluate. We aim to introduce these advancements to the neurosurgical audience while suggesting major potential roadblocks to their safe and effective translation. Unlike the previous generation of machine learning in neurosurgery, the safe translation of recent advancements will be contingent on neurosurgeons' involvement in model development and validation.

European Stroke Organisation (ESO) Guidelines on Moyamoya angiopathy: Endorsed by Vascular European Reference Network (VASCERN)

The European Stroke Organisation (ESO) guidelines on Moyamoya Angiopathy (MMA), developed according to ESO standard operating procedure and Grading of Recommendations, Assessment, Development and Evaluation (GRADE) methodology, were compiled to assist clinicians in managing patients with MMA in their decision making. A working group involving neurologists, neurosurgeons, a geneticist and methodologists identified nine relevant clinical questions, performed systematic literature reviews and, whenever possible, meta-analyses. Quality assessment of the available evidence was made with specific recommendations. In the absence of sufficient evidence to provide recommendations, Expert Consensus Statements were formulated. Based on low quality evidence from one RCT, we recommend direct bypass surgery in adult patients with haemorrhagic presentation. For ischaemic adult patients and children, we suggest revascularization surgery using direct or combined technique rather than indirect, in the presence of haemodynamic impairment and with an interval of 6–12 weeks between the last cerebrovascular event and surgery. In the absence of robust trial, an Expert Consensus was reached recommending long-term antiplatelet therapy in non-haemorrhagic MMA, as it may reduce risk of embolic stroke. We also agreed on the utility of performing pre- and post- operative haemodynamic and posterior cerebral artery assessment. There were insufficient data to recommend systematic variant screening of RNF213 p.R4810K. Additionally, we suggest that long-term MMA neuroimaging follow up may guide therapeutic decision making by assessing the disease progression. We believe that this guideline, which is the first comprehensive European guideline on MMA management using GRADE methods will assist clinicians to choose the most effective management strategy for MMA.

Surgery for brain metastases: radiooncology scores predict survival-score index for radiosurgery, graded prognostic assessment, recursive partitioning analysis

BACKGROUND

Radiooncological scores are used to stratify patients for radiation therapy. We assessed their ability to predict overall survival (OS) in patients undergoing surgery for metastatic brain disease.

METHODS

We performed a post-hoc single-center analysis of 175 patients, prospectively enrolled in the MetastaSys study data. Score index of radiosurgery (SIR), graded prognostic assessment (GPA), and recursive partitioning analysis (RPA) were assessed. All scores consider age, systemic disease, and performance status prior to surgery. Furthermore, GPA and SIR include the number of intracranial lesions while SIR additionally requires metastatic lesion volume. Predictive values for case fatality at 1 year after surgery were compared among scoring systems.

RESULTS

All scores produced accurate reflections on OS after surgery (p ≤ 0.003). Median survival was 21-24 weeks in patients scored in the unfavorable cohorts, respectively. In cohorts with favorable scores, median survival ranged from 42 to 60 weeks. Favorable SIR was associated with a hazard ratio (HR) of 0.44 [0.29, 0.66] for death within 1 year. For GPA, the HR amounted to 0.44 [0.25, 0.75], while RPA had a HR of 0.30 [0.14, 0.63]. Overall test performance was highest for the SIR.

CONCLUSIONS

All scores proved useful in predicting OS. Considering our data, we recommend using the SIR for preoperative prognostic evaluation and counseling.

The Clinical Frailty Scale as predictor of overall survival after resection of high-grade glioma

BACKGROUND

The Clinical Frailty Scale (CFS) describes the general level of fitness or frailty and is widely used in geriatric medicine, intensive care and orthopaedic surgery. This study was conducted to analyze, whether CFS could be used for patients with high-grade glioma.

METHODS

Patients harboring high-grade gliomas, undergoing first resection at our center between 2015 and 2020 were retrospectively evaluated. Patients' performance was assessed using the Rockwood Clinical Frailty Scale and the Karnofsky Performance Scale (KPS) preoperatively and 3-6 months postoperatively.

RESULTS

289 patients were included. Pre- as well as postoperative median frailty was 3 CFS points (IqR 2-4) corresponding to "managing well". CFS strongly correlated with KPS preoperatively (r = - 0.85; p < 0.001) and at the 3-6 months follow-up (r = - 0.90; p < 0.001). The reduction of overall survival (OS) was 54% per point of CFS preoperatively (HR 1.54, CI 95% 1.38-1.70; p < 0.001) and 58% at the follow-up (HR 1.58, CI 95% 1.41-1.78; p < 0.001), comparable to KPS. Patients with IDH mutation showed significantly better preoperative and follow-up CFS and KPS (p < 0.05). Age and performance scores correlated only mildly with each other (r = 0.21…0.35; p < 0.01), but independently predicted OS (p < 0.001 each).

CONCLUSION

CFS seems to be a reliable tool for functional assessment of patients suffering from high-grade glioma. CFS includes non-cancer related aspects and therefore is a contemporary approach for patient evaluation. Its projection of survival can be equally estimated before and after surgery. IDH-mutation caused longer survival and higher functionality.

Machine learning-based clinical outcome prediction in surgery for acromegaly

PURPOSE

Biochemical remission (BR), gross total resection (GTR), and intraoperative cerebrospinal fluid (CSF) leaks are important metrics in transsphenoidal surgery for acromegaly, and prediction of their likelihood using machine learning would be clinically advantageous. We aim to develop and externally validate clinical prediction models for outcomes after transsphenoidal surgery for acromegaly.

METHODS

Using data from two registries, we develop and externally validate machine learning models for GTR, BR, and CSF leaks after endoscopic transsphenoidal surgery in acromegalic patients. For the model development a registry from Bologna, Italy was used. External validation was then performed using data from Zurich, Switzerland. Gender, age, prior surgery, as well as Hardy and Knosp classification were used as input features. Discrimination and calibration metrics were assessed.

RESULTS

The derivation cohort consisted of 307 patients (43.3% male; mean [SD] age, 47.2 [12.7] years). GTR was achieved in 226 (73.6%) and BR in 245 (79.8%) patients. In the external validation cohort with 46 patients, 31 (75.6%) achieved GTR and 31 (77.5%) achieved BR. Area under the curve (AUC) at external validation was 0.75 (95% confidence interval: 0.59-0.88) for GTR, 0.63 (0.40-0.82) for BR, as well as 0.77 (0.62-0.91) for intraoperative CSF leaks. While prior surgery was the most important variable for prediction of GTR, age, and Hardy grading contributed most to the predictions of BR and CSF leaks, respectively.

CONCLUSIONS

Gross total resection, biochemical remission, and CSF leaks remain hard to predict, but machine learning offers potential in helping to tailor surgical therapy. We demonstrate the feasibility of developing and externally validating clinical prediction models for these outcomes after surgery for acromegaly and lay the groundwork for development of a multicenter model with more robust generalization.

Preoperative nonmedical predictors of functional impairment after brain tumor surgery

PURPOSE

To identify the preoperative nonmedical predictors of functional impairment after brain tumor surgery.

METHODS

Patients were evaluated before brain tumor surgery and after 3 months. The cognitive evaluation included MOCA for the general cognitive status, TMT for attention and executive functions, ROWL-IR and ROWL-DR for memory, and the F-A-S for verbal fluency. Anxiety, depression, social support, resilience, personality, disability, and quality of life were evaluated with the following patient-reported outcome measures (PROMs): HADS, OSS-3, RS-14, TIPI, WHODAS-12, and EORTC-QLQ C30. Functional status was measured with KPS. Regression analyses were performed to identify preoperative nonmedical predictors of functional impairment; PROMs and cognitive tests were compared with the normative values.

RESULTS

A total of 149 patients were enrolled (64 glioma; 85 meningioma). Increasing age, lower education, higher disability, and lower ROWL-DR scores were predictors of functional impairment in glioma patients while higher TMT scores and disability were predictors in meningioma patients. In multiple regression, only a worse performance in TMT remains a predictor in meningioma patients. Cognitive tests were not significantly worse than normative values, while psychosocial functioning was impaired.

CONCLUSION

TMT could be used in the preoperative evaluation and as a potential predictor in the research field on outcome predictors. Psychosocial functioning should be studied further and considered in a clinical context to identify who need major support and to plan tailored interventions.